Conventionally, there is a known technology in which a commodity contained in image data is recognized (identified) by comparing the data representing the shape and the like of a standard commodity with the data extracted from the image data. In such a technology, data of commodities in a plurality of categories are registered in advance and then it can be possible to increase the number of commodities which are recognized in correspondence with the number of commodities registered. However, if the number of the registered data is increased, the time taken for the data comparison is increased as well, which may lead to a low recognition efficiency. Further, if the number of the registered data is increased, the frequency of the occurrence of the incorrect recognition may also increase.